{ Scanned and converted to text by Jed Margolin. Does not have drawings.
The original document is the controlling document. }
0061989 E.I. Monthly No. EI7704021861 E.I. Yearly No. EI77003363
Title: SOME NAVIGATIONAL CONCEPTS FOR REMOTELY PILOTED VEHICLES
Author: Lyons, J.W.; Bannister, J.D.; Brown, J.G.
Corporate Source: Hawker Siddeley Aviat Ltd, Brough, North Humberside, Engl
Source: AGARD Conference Proceedings n 176 Aug 1976 Medium Accuracy Low Cost
Navig at Avionics Panel Tech Meet, Sandfjord, Norw, Sep 8-12 1975 Pap 15 p
Publication year: 1976
CODEN: AGCPAV ISSN: 0549-7191
Language: English
629.13 A063A4
Lyons 5-1
SOME NAVIGATIONAL CONCEPTS FOR REMOTELY PILOTED VEHICLES
J.W. Lyons, J.D. Bannister, J.G. Brown.
Abstract
--------
This paper discusses methods by which the navigation function for Remotely
Piloted Vehicles (RPVs) can be achieved without the need for complex
specialised navigation equipment. The objective is to make use of equipment
normally carried for RPV operation to supplement a simple dead reckoning
navigation system. In this way significant improvements in navigation
capability can be achieved with little or no added complexity in the vehicle
itself. The additional processing is carried out at the control centre where
restrictions on equipment size and cost are not so prohibitive. Both a two-way
data link and a forward looking electro-optical sensor are highly desirable
RPV facilities and these are on-board equipments that can be adapted to
provide additional information at the ground-based or airborne control station
for vehicle position updating.
The paper discusses techniques varying from the use of the data link to
provide range-bearing navigation to map matching using reconnaissance sensors
or a forward looking sensor picture. Use can also be made of an on-board laser
to provide range-to-terrain measurements which, when correlated with a
computer stored map, enables the RPV position to be continuously updated.
Results of simulation studies which have been carried out to validate the
techniques and provide an estimate of the accuracies that may be achieved are
presented.
{ Note that the original uses Greek symbols which I cannot enter in their
original form. Therefore, it has been necessary to spell them out.}
NOMENCLATURE
------------
sigma_RPV = Position error of RPV
sigma_R = Range error of DME system
sigma_psi = Bearing error of Data Link
R = Range of RPV from relay aircraft
sigma_A = Navigation error of control or relay vehicle
Rn = Range of RPV at the nth sample
psi_n = Azimuth angle of RPV at the nth sample
delta_t = Time between data samples
VR = Velocity of relay vehicle
theta = Heading of RPV
RC = Range of RPV from the bisector of the relay station base line
psi_C = Bearing of RPV from the bisector of the relay station base line
D = Distance between the relay stations forming the base line
RIP = Range from RPV to Identification Point
h = Height of RPV above Identification Point
theta_IP = Downlook angle from RPV to Identification Point
theta_L = Laser depression angle
phi_i = Laser azimuth angle
RHi = Horizontal range from RPV to laser/terrain intersection point
delta_Hi = Heigth difference between terrain at RPV and at laser/terrain
intersection point
epsilon_i = Error in actual/predicted terrain height
1. Introduction
---------------
In recent years the ever increasing cost and complexity of manned aircraft
for operation in a battlefield environment has led to a re-appraisal of the
use of Remotely Piloted Vehicles (RPVs) for certain types of missions. For
high attrition situations in which aircrew are at risk the use of expendable
or limited life vehicles is attractive. Provided the vehicle controllers are
{provided?} with the necessary guidance and control information, the RPV can
possess an operational flexibility comparable with that of a manned aircraft.
The roles most suited to a battlefield RPV are:
i) Target Marking
ii) Reconnaissance
iii) ECM
The penetration of the RPV beyond the Forward Edge of the Battle Area
(FEBA) necessitates the use of a relay station located such that its altitude
is adequate to maintain radio contact with the RPV while
==============================================================================
Lyons 5-2
its position is such as to be out of range of SAMs. The relay may be either a
stationary platform or a patrolling aircraft. In the latter case, the
controller can be located in the aircraft. More usual is the use of a ground
control station.
The RPB should be as small as possible compatible with the above mission
tasks and this means restricting the complexity of the onboard avionics.
Although equipment such as forward looking and reconnaissance sensors, a data
link and possibly a laser are of necessity located on the vehicle, the
navigation and guidance equipment can be largely accomodated on the relay
vehicle or at the ground station. The sensors already on board the RPV can be
used to provide a navigational facility which can supplement a simple modest
accuracy system such as a compass/air data unit. The basic airborne system
would provide sufficient information for general flying of the RPV, i.e.
heading, velocity and a rough measure of position, while the additional
sensors can be used to provide an accurate measure of present RPV position.
This philosophy is adopted here and the paper presents a number of alternative
techniques whereby, depending on the particular situation, one or more of the
above items form part of the overall navigation system.
Firstly, the data link is required to maintain a constant or regular
periodic contact with the RPV by means of a narrow beam-width microwave link,
hence a tracking facility must already exist on the relay vehicle providing
RPV bearing information. Range information can be provided by means of a
responsive transponder similar to an IFF system utilizing the same vehicle
antennas.
Secondly, update facilities can be provided by means of either a real time
forward looking or vertical reconnaissance image used in conjunction with a
moving map display.
A third possibility makes use of the ranging laser used for target marking
purposes. En route to and from the target area, range-to-terrain measurements
can be transmitted over the data link to the control station. This data can
then be correlated with a computer stored map to determine the most likely RPV
position.
The adoption of one or more of the above techniques leads to a significant
improvement in navigational accuracy with little or no additional complexity
in the vehicle itself.
2. RADIO NAVIGATION USING A DATA LINK
-------------------------------------
The data link forms the life line of communication between the RPV and the
control station. It is the means by which guidance signals to the RPV are
transmitted and video signals received. Because of the need for wideband
transmissions of video signals (typically 5 MHz) and the desirability of
narrow beamwidth, low side-lobe antennas for good anti-jamming capability,
microwave frequencies are generally employed. This limits RPV operation to
line of sight communication and hence may necessitate the use of airborne
relay stations. A possible operational situation is shown in Fig. 1. In
practice there may well be more than one relay station and RPV. It is
envisaged that the relay station will stand back from the FEBA, out of direct
range of ground-to-air weapons. This does not however prevent the enemy making
use of either ground or airborne jammers to illuminate the relay vehicle,
thereby reducing the effective signal-to-noise ratio of the signals received
from the RPV. Two situations can be distinguished, one in which the relative
relay - RPV geometry is such that the jamming signals are received by the
relay antenna mainlobe, in which case the signal-to-noise ratio is low. The
second situation relates more to large lateral separations of jammers and the
RPV in which case jamming signals enter the relay antenna via the side-lobes.
In such cases, the signal-to-noise ratio may not be significantly degraded and
unimpaired operations can continue.
When the effects of enemy ECM can be neglected, i.e. the relay station
remaining in contact with the RPV, angular information is directly available
from the data link antenna and range can be derived using conventional DME
techniques. Thus the position of the RPV relative to the relay station can be
reasonably well defined. For absolute location of the RPV, clearly the
position of the relay vehicle needs to be defined. In the case of tethered
platforms this is no problem but for patrolling aircraft or hovering vehicles
the error of the relay vehicle navigation system has also to be taken into
account. An overall error can be estimated from the following equation.
sigma_RPV = sqrt( sigma_R**2 + sigma_A**2 + (R**2)*sigma_psi**2 ) -(1)
Typical results are presented in Fig. 2.
Perhaps of more importance is the dynamic problem of guiding the RPV to a
given position. For this case it is desirable to have a good knowledge of the
RPV heading and velocity as well as its present position and best results are
obtained by using both on-board and remote guidance equipment. For example,
estimates of heading and velocity provided by the compass/air data system can
be compared with time dependent range and bearing data derived from the data
link to obtain improved estimates of RPV position, velocity and heading.
Figure 3 shows the geometry relevant to a 3 point moving window tracking
technique. The heading of the RPV can be written in functional form as
theta = f( R(n-1,n,n+1),psi(n-1,n,n+1)/delta_t,VR ) -(2)
This generally requires more processing effort than the determination of range
or velocity. For tethered or hovering relay vehicles VR is clearly zero in the
above equation. Since the on-board and remote systems use independent data the
results are best combined using a statistical filter. The simplest approach is
to use a least squares technique (see Reference 1). Alternatively, an
integrated filtering method as described in Reference 2 may be employed. This
latter paper suggests a significant improvement in navigational accuracies by
employing filtering techniques.
In ECM environments, range information to the RPV cannot be guaranteed
though it is likely that bearing information can still be derived. To estimate
the RPV position in such circumstances, use can be made of the possible
multiplicity of relay stations. From known locations of the relay vehicles,
cross bearing fixes on the RPV of interest can be achieved. This is a well
known location technique, both for air and marine applications. A detailed
analysis of the method is given in Reference 3. For the present
==============================================================================
Lyons 5-3
analysis a more useful expression for position accuracy is
sigma_RPV = sigma_phi * sqrt(2) * {[(RC**2)-(D**2)/4)]**(1/2)} -(3)
/[RD * cos(phi_C)]
* { [RC**2 - D**2 /4]**2 - [RC**2 * sin(phi_C)**2] }**(1/2)
Results derived from equation 3 are plotted in Fig. 4. It can be shown from
the above expression that the best accuracy is achieved when phi_C = 0 and
RC/D = 0.3536. Thus for good accuracy using this technique, the separation
between the relay stations should be large compared with the penetration of
the RPV beyond the FEBA. To determine the overall RPV position, the additional
effect of relay station position accuracy must also be taken into account.
3. MAP MATCHING
---------------
So far we have considered on-board dead reckoning and remote radio
navigation techniques. The main problem with these techniques is that the
position accuracy is either time or range dependent and so additional methods
of updating vehicle positions are necessary. A number of techniques are
available for an RPV. For reconnaissance vehicles having real time sensors,
the problem is relatively straight-forward. The use of either Side Looking
Radar (SLR) or Infra Red Line Scan (IRLS) systems means that effectively a map
is generated while the sensor is operating. The resulting video signal
transmitted to the control station thus provides a method whereby the RPV
position can be readily located.
One system widely employed for displaying aircraft navigational information
is the projected moving map display and a similar technique can be employed by
the RPV control station. Current map systems have the additional facility of
being able to combine an electronic display with the moving map and Reference
4 discusses some of the latest developments in this field. Making use of this
principle, it may be possible to project the sensor image onto the map and
determine the RPV position by matching the two images. Fig. 5 shows the
principles of the combined map/sensor display projection system.
In practice it is envisaged that the RPV reconnaissance sensor image will
be monitored on a TV display. The use of digital scan converters allows a
number of alternative display presentations (see Reference 5). Perhaps the
most convenient display mode for the present application is the rolling map or
"passing scene" technique where a new line is added to the top of the display
and the scene is shifted slowly downwards.
When likely update features are seen (e.g. rivers, crossroads, distintive
manmade objects) the frame is frozen, a transfer button is initiated and the
digitally stored frame is projected via the map system. The map is then moved
laterally to align with the projected image. When the alignment is judged
adequate an accept button is pressed and the present position co-ordinates of
the RPV updated, taking into account the elapsed time for updating actions. A
possible arrangement of operator console is shown in Fig. 6. Control of the
image pictures and map matching facility is achieved through the use of a
joystick control. Some simulated results of this update technique are shown in
Fig. 7. These results make use of SLR imagery.
When the RPV has only real time forward-looking sensors, use can still be
made of the transmitted image to provide a navigational update facility.
However, in order to create the correct perspective map-like projection,
appropriate transformation of the image is necessary. In photogrammetrical
language this is termed rectification though the appropriate term in
perspective art is anamorphic projection. The principle involved is shown in
Fig. 8. The received forward looking image may be co-ordinate transformed
either by optical techniques utilizing anamorphic lens system or
electronically by means of the scan converter or projection CRT sweep
circuitry. Since the image already exists in electrical form, the electronic
transformation techniques are probably most suitable. The map type image
projected onto the display is now trapeziodal in shape because of the
transformation. Major features on the map can again be aligned as described
above. In practice several factors combine to make the task more difficult
than for the vertical sensor case:
i) varying resolution, contrast and intensity across the display.
ii) distortion due to undulation of the terrain.
iii) the wildly exagerated size of trees, hedges, buildings ets.
Hence an alternative simpler update technique is proposed for this situation.
With a forward looking sensor display it is possible to mark objects
electronically with a joystick controlled marker symbol; this is standard HUD
technology. The electronics can be arranged such that having frozen a suitable
image and marked an identifiable point on it, a marker symbol appears on the
projected map. Also the field-of-view of the sensor, as projected in the
horizontal plane, is superimposed on the map as a "bright up" presentation so
that the orientation of the sensor view is clearly seen. The same joystick is
now used to align the map with the marker. To ensure correct alignment at
least two identification points (IPs) are required on any given image,
preferably three or four. In a conventional airborne situation the task of
marking a target on a display is not easy and may take several seconds. For
the situation described above, however, the problem is one of marking chosen
objects on a frozen image in a shirt sleeve environment and hence this aspect
of the navigation problem is not considered too difficult.
Fig. 9 shows some simulated results of the above update technique. The
effect of the bright area is clearly seen in relationship to marked targets.
4. TERRAIN MAP CORRELATION
--------------------------
Reconnaissance or forward looking sensors provide a convenient method of
updating the navigation system. However, these sensors require a large data
link bandwidth to transmit the video picture to the control centre and hence
are vulnerable to ECM. Reduction of the video bandwith reduces the effect of
ECM but with a consequent degradation of picture resolution. Hence an
alternative method of updating the navigation system is desirable. The method
to be described uses ranging measurements made by the
==============================================================================
Lyons 5-4
Laser and compares these with corresponding ranges obtained from a
representation of the terrain stored in a computer at the control centre. The
data link bandwidth required to transmit the laser ranges is very small and
hence is correspondingly less susceptible to interference by ECM.
Basically the technique depends on an adequate representation of the
terrain over which it is intended to fly the RPV. The terrain is stored as a
series of height ordinates obtained from a map of the relevant area and these
are used to construct a computer model of the terrain (Fig. 10). The initial
effort in producing this data base from the map is considerable but for a
given area it is a 'once-only' task. A simulation of the RPV flight path at
the control centre then allows laser range to be calculated for each RPV
position and a comparison made with actual ranging measurements. A series of
positions and headings around the expected values (and limited in deviation
from these expected values by estimated navigation errors) are also tested
against the actual measurements and the best position and heading for the RPV
found.
For a 2-D simulation, where it is only necessary to determine the
alongtrack position of the RPV, it has been found that a minimum of three
measurements (2 laser - altimeter) are necessary to give a reliable indication
of position, while for a 3-D simulation at least four measurements (3 laser -
altimeter) are required. These conclusions are based on error-free
simulations. However, when errors are taken into account it has been found
necessary to considerably increase the number of measurements to effectively
smooth out the errors. Apart from the errors involved in the actual laser
measurements the accuracy of terrain representation has a considerable
influence on the feasibility of the method. In addition, the technique is
ineffective over the sea or over flat, featureless terrain. Nevertheless, by
combining this method with those described previously, an effective navigation
system is offered without the necessity for specialised navigation equipment.
The method has been demonstrated using a computer simulation of both the
laser range measurement and range matching processes, bearing in mind that the
latter should not simply be a reversal of the former as this would neglect the
"real world" errors caused by imperfect representation of the terrain. The
simulation of the matching process is precisely the process that is required
to be carried out at the control centre, while the simulation of the laser
measurement is an attempt to predict the results of actual measurements made
from the vehicle during flight. Hence careful representation of the terrain
has been used for measurement simulation with terrain data points spaced 100m
apart on a rectangular grid.
The range as seen by the laser is calculated by taking a section through
the terrain in the direction in which the laser is pointing. Assuming a
knowledge of the RPV height above the terrain h (from a radio altimeter) and
the laser beam depression angle theta_L, the horizontal range RH and
incremental height delta_Hi of the laser/terrain intersection point Y,
relative to the RPV position X, can be calculated (Fig. 11). The following
data is then transmitted from the RPV to the control centre:
i) height differences delta_H1 ... delta_Hi ... delta_Hn
ii) horizontal ranges RH1 ........ RHi ......... RHn
iii) laser azimuth angles theta_1 .... theta_i ..... theta_n
From a knowledge of RPV velocity and heading and an estimate of likely
navigation errors, the current RPV position can be predicted together with a
circle of possible error (Fig. 12). A search can therefore be made within the
circle to determine the most likely RPV position. For each position
considered, the terrain height H is known from the model and at range RH and
bearing phi_i from that position the expected terrain height is given by
H + delta_Hi. This is compared with the actual terrain height at that point
(as stored by the model) to give an error epsilon_i. By considering each RHi
and phi_i (i = 1 to n) an RMS error is obtained for each position, and the
position with minimum error gives the most likely RPV position.
5. NAVIGATION ACCURACIES
------------------------
In this section of the paper an attempt will be made to compare the
navigation accuracies attainable from the various techniques previously
discussed.
For the basic on-board system comprising a magnetic compass and air data
unit, the following accuracies are predicted based on currently available
equipment:
heading 1 (degree) standard deviation
velocity 2% standard deviation
This gives a position accuracy of approximately 2% distance gone. However,
a major source of error will be due to wind; although a correction can be
applied, an uncertainty in wind speed of the order of 5 m/s is not
unreasonable. Assuming an RPV velocity 200 m/s this represents 2 1/2 % giving
a resultant position accuracy of the order of 3 1/2 % distance gone.
Range-bearing techniques have been used for many years as exemplified by
TACAN/DME navigation. When using ground beacons a major source of error is
multipath propagation which gives rise to large errors in estimating the
bearing to a station. However the modern systems which use airborne beacons
overcome this pr???? ??? [problem?] this is the situation which exists
when considering RPVs.
Clearly target bearing estimation from the relay vehicle is a major
contributor to RPV location accuracy. Since microwave frequencies, perhaps at
X-band, coupled with monopulse determination techniques are employed in the
relay vehicle, good angular estimates of the RPV bearing are available. Final
figures are dependent on antenna size, frequency of operation and signal-to-
noise ratio. It is considered that at least 1(degree) standard deviation
should be readily attainable in a practical system. From Fig. 2 it is seen
that this gives a typical RPV position error better than 2 km standard
deviation at 100 km range. The ultimate short range accuracy is clearly
dependent on the accuracy of the relay vehicle navigation system.
=============================================================================
Lyons 5-5
When jamming environments are such that perhaps only bearing information is
available to the relay vehicles the cross bearing fix principle utilising
multiple relay vehicles remains a possibility for RPV position fixing. Fig. 4
shows the accuracy function on a relative scale and clearly indicates the
position dependent accuracy effect. To utilise this technique successfully in
a practical situation, it is necessary to carefully select the patrol station
positions for the relay vehicles relative to the battlefield.
Taking the 50% accuracy contour as a guide to the area of utility of the
technique, this corresponds to a distance from the baseline bi-sector roughly
equal to the relay station separation. If we therefore envisage RPV operations
out to 100 km from the relay, the relay stations should be located 100 km from
each other. At this separation, with a bearing accuracy estimation of
1(degree) standard deviation the RPV can be located to a maximum accuracy of
1.6 km standard deviation. Combining this with a typical relay vehicle
position accuracy of 0.5 km raises this figure by less than 0.1 km.
Navigation updating using a real time picture from a vertical
reconnaissance sensor provides a very accurate means of position fixing.
Fig. 7 shows some simulated results based on SLR imagery. The picture quality
of these radars is seen to be more than adequate to identify the main
geographical terrain and man made features. In the example shown, the river
bank provides a good map matching feature. Fig. 7a shows some degree of
misalignment of the map and radar image. In Fig. 7b the two are aligned. Some
errors are present due to the scale compression effect at ranges close to the
RPV and this is reflected in map projection distortion. Even without further
video processing to correct this effect, it is considered that a location
accuracy of 0.2 km is attainable.
When using a forward looking sensor for map matching the useful range of
the sensor is limited to ~3 km, hence the matching will be done over a small
area and a larger scale map can be used (of Figs. 7 and 9). This, together
with the fact that considerable detail will be visible in the foreground of
the display, makes the matching task easier allowing a match to within say
100 m. Unfortunately various system errors can produce incorrect
transformation of the display and result in significant position errors. The
sources of error and their effects are the same irrespective of whether a full
display transformation technique is being used or only marked identification
points.
Across track errors should be small since the only error is that due to
marking the display in azimuth. Display marking should be possible to within
+/- 2% full scale, allowing for both operator and marker system errors. For a
30(degree) FOV sensor this corresponds to an angular error of 10 m rads.
Display points of interest are expected to be at ranges between 1 and 2 km and
for accurate across track matching a near and a far point should be chosen.
This will give sensor heading to within 30 m rads and across track errors
<40 m, i.e. the matching is the biggest source of error.
Along track errors can be much greater. The range to an identification
point is given by
RIP = h/tan(theta_IP)
where h is the height of the RPV above the IP
theta_IP is the downlook angle from RPV to the IP
The most significant sources of error in determining RIP with typical values
for standard deviation, are
i) Uncertainty in RPV altitude ~3 m in 150 m i.e. 2% h
ii) Undulating terrain. The effect of undulating terrain is exactly the
same as variations in RPV altitude. Variations ~ +/- 20 m are
expected, i.e. 13% h.
iii) Display marking. Errors in marking the display in elevation are again
estimated at +/- 2% full scale. For a 20(degree) vertical FOV this is
8 m rad.
iv) Uncertainty in sensor attitude. The accuracy with which the sensor
attitude is known in elevation is dependent on the equipment fit in
the RPV. A value of 2 m rad is assumed. If the attitude is not known
to this accuracy an estimate can probably be made from the position
of the horizon.
For identification points at a nominal range of 1.5 km the above factors give
the following independent errors
i) 30 m ii) 200 m iii) 100 m iv) 25 m ???
The combined effects of these errors and the basic matching error is 230 m.
As yet it has not been possible to quantify the navigation accuracy that
could be achieved by the laser/terrain correlation system. It is a function of
the terrain used and the accuracy of terrain representation. Preliminary
results of the simulation described previously are available with the effects
of errors in
- laser beam depression angle ( 2 m rad, 1 sigma )
- laser range measurement ( 6 m, 1 sigma )
- radio height measurement ( 3 m, 1 sigma )
- terrain height reprentation ( ~3 m, 1 sigma )
represented. These results suggest that the technique is viable. Nevertheless
the search technique used to obtain these results was very much simplified;
for each navigation attempt the true vehicle position was presented to the
system along with numerous points in the search area. In practice, the true
position would not be available and some degradation in results would then be
expected.
Further work is required to ascertain the relation between navigation
accuracy and errors in terrain representation. However, since it appears that
terrain representation is an important part of the concept terrain data taken
directly from stereoscopic photographs should yield considerable improvement
over data taken from maps. Also careful consideration is required of the
optimum search technique which should be used in practice.
Lyons 5-6
6. CONCLUSIONS
--------------
A navigation concept has been presented whereby a good navigation accuracy
(down to 1/4 km) can be realised for an RPV with the minimum of on-board
equipment. Table 1 summarises the accuracies of the various techniques
available. It is proposed that several of these be incorporated into the
overall RPV control and guidance system so that the controller can select the
one most suitable for a given situation.
When a wide bandwidth data link can be maintained the map matching
technique using SLR or IRLS offers the simplest and most accurate solution
with the forward looking sensor as a good alternative. It does, however,
impose a large workload on the controller since, depending on the accuracy of
the basic on-board system, the updating needs to be performed every few
minutes. A separate navigator is therefore envisaged, keeping track of several
RPVs. Electronic devices which are currently being developed to perform area
correlation for automatic electro-optical tracking may lead to automation of
the matching task in the future.
Where the data link is limited in bandwidth the laser/terrain correlation
technique should give good accuracy and the process could be completely
automated to provide a continuous indication of RPV position. Disadvantages of
the system are the large amount of data storage and computation necessary at
the control centre, the development work required to produce an operational
system and the unsuitability of the system over featureless terrain.
Alternatively recourse can be made to a system based on measurements made
from the relay stations. These are well established techniques offering good
accuracy at short ranges and modest accuracy at long ranges. Again a
completely automatic system is possible.
In the event of a total failure of the RPV control/guidance link, the
on-board system would be adequate to allow the RPV to navigate itself back to
a pre-defined recovery area.
7. ACKNOWLEDGEMENTS
-------------------
The authors acknowledge the help given by H. G. Loftus and his colleagues
during the preparation of the photographic material for this paper. Permission
to publish the paper is by courtesy of Hawker Siddeley Aviation Limited. The
opinions however are entirely those of the authors.
8. REFERENCES
--------------
1. Ramsayer K, 'Integrated Navigation by Least Square Adjustment',
AGARD C.P. No. 54, September 1969. P 18-1.
2. Hemesath, N.B., 'Optimal and Suboptimal Velocity - Aiding for VOR/DME
Systems', AIAA Journal Vol. 10, No. 1, January 1972, P. 24
3. Stansfield, R.G., 'Statistical Theory of D.F. Fixing', Proc. I.E.E. 94,
Pt. III a, No. 15, 1947.
4. Aspin, W.M. 'Comed - A Combined Display including a Full Electronic
Facility and a Topographical Moving Map Display', AGARD C.P.P. No. 167,
April 1975, P.30-1.
5. Slocum, G.K. 'Digital Scan Converters in Airborne Display Systems',
AGARD C.P.P. No. 167, April 1975, P.23-1.
Table 1
-------
Comparison of RPV Navigation Techniques
---------------------------------------
CONTINUOUS NAVIGATION:
Technique Accuracy-km (1 sigma) Comment
------------------- --------------------- ----------------------------
Compass/Air-Data 3.5 after 100 km 3 1/2 % Distance gone
Basic On-Board System Depends on wind estimates
Range-Bearing 1.8 at 100 km range 1(degree) Bearing accuracy
from Relay Station
Cross Bearing Fix 1.6 at 100 km range 1(degree) Bearing accuracy
from Relay Stations 100 km baseline
Laser Ranger-Terrain 0.5 Depends on the accuracy of
Map Correlation the terrain representation
UPDATE TECHNIQUES:
Technique Accuracy-km (1 sigma) Comment
------------------- --------------------- ----------------------------
Map Matching with 0.2 Accuracy limited by
Recce Sensor display system
Map Matching with 0.23 As above. Additional
Forward Looking Sensor errors due to display marking
etc. Altitude 150 m.
Figures
-------
Fig. 1 RPV Operational Situation
Fig. 2 Accuracy of DME System
Fig. 3 Moving Window Tracking Technique
Fig. 4 Plot of results derived from equation 3
Fig. 5 Combined Moving Map/CRT Display
Fig. 6 RPV Controllers Console
Fig. 7 Simulated SLR/MAP Update System
Fig. 8 Forward Looking Sensor Co-ordinate Transformation
Fig. 9 Simulation of Marked Forward Looking Display/Map Update System
Fig. 10 Terrain Model
Fig. 11 Terrain Section
Fig. 12 Terrain Correlation Search
Discussion
-----------
D Halliwell, Decca Systems Study and Management Division, UK
Using the terrain map correlation method, are three ranges really able to give
an unique position? There are probably many solutions in each case, only one
of which is correct. After a false reset the true position may be outside the
area of uncertainty for the next fix. Have your simulations shown any tendency
to this effect?
J W Lyons, HSA, UK
For an error-free system three range measurements and radio height will in
general be adequate to give an unique position within a limited area, though
it is possible to conceive terrain configurations where this would not hold.
The method will not work over flat featureless terrain. Also, in a real-world
system, errors will be present and further range measurements will be
necessary to smooth the effects of these. For convenience and to avoid a
cluttered presentation only three measurements were illustrated in Fig. 12.
The area of uncertainty for the next fix depends on errors associated with the
estimation of present position. However, when an update is attempted, a
confidence level can be estimated based on how well the range measurements fit
the stored terrain model. Only when a high confidence level is achieved is an
update accepted.
C T J Jessop, Sperry Gyroscope Company, UK
To achieve the fix accuracies quoted what horizontal datum accuracy, in pitch
and roll, is assumed for forward and sideways looking laser and radar sensors;
and could these in fact approach inertial navigation system accuracy levels?
J D Bannister, HSA, UK
For the small laser beam depression angles assumed, the system is relatively
insensitive to small changes in pitch and roll angles. The paper illustrates,
in Fig. 11, that it is the horizontal range, RH, which is used for the
correlation process. The error in RH will be small. However the question then
arises as to the change in terrain height over the distance associated with
the error in RH. This will depend very much on the nature of the terrain being
overflown. The accuracy of the pitch and roll information thus determines the
type of terrain over which the method provides a useful update facility. Also
it should be borne in mind that the smoothing effect of taking a number of
measurements is very powerful.